The present invention relates to an analyzing unit, an inspecting system and a manufacturing method using the inspecting system, which are applied to a production line used in the manufacture of electronic devices and the like; and, more particularly, the invention relates to an analyzing unit, an inspecting system and a manufacturing method using the inspecting system, which efficiently classify images obtained as a result of inspection of devices being manufactured, to thereby shorten the analysis time required to detect a failure on the production line and to enable enhancement of manufacturing efficiency and a rapid yield ramp-up.
An electronic device, typically a semiconductor, is formed by repeating a plurality of processing steps, including exposure, development, etching and the like, on a wafer substrate. On the other hand, in certain ones of the plurality of processing steps, information indicating the position, size, number, category, etc. of foreign matter which has adhered to the wafer, as well as appearance failures, and information indicating the processing dimensions of workpiece (hereinafter generally referred to as defects) are collected by a foreign matter inspecting apparatus, an optical visual inspection apparatus, and an inspecting apparatus, such as a SEM, if necessary. In Monthly Publication, “Semiconductor World”, 1996.8, pp 88, 99 and 102, it is indicated that all of the inspecting data are normally sent from the inspecting apparatus to an analyzing system through a network for control and analysis of the production.
Further, in the plurality of processing steps, an electronic microscope or the like is used to specify the cause of an occurrence of defects to obtain a defect image and perform a classifying operation on the basis of the shapes and sizes of the actual defects. This classifying operation is carried out visually on the screen of a personal computer or the like, whereby defect images are sampled on the basis of the size, the shape and the like of the defects, and then the defects are classified into groups of similar defects. It is noted that for the defects to be sampled, several numbers of defects per wafer are manually determined, referring to the distribution of defects as a wafer map. Recently, several hundreds of defect images per hour have been automatically obtained by an automatic defect review (ADR) of defect images. Anyway, the number of image samples being handled has tended to increase.